SPECTRUM ANALYSIS - THE WAVE ANALYSER

Transcription

1 SPECTRUM ANALYSIS - THE WAVE ANALYSER PREPARATION the SPECTRUM ANALYSER principle of operation...34 practical variation...35 the WAVE ANALYSER the model...36 spectrum measurement - single component input...36 spectral measurement - two component input...37 practical considerations precautions...38 searching methods...38 where?...38 how large?...38 EXPERIMENT SPECTRUM UTILITIES module VCO fine tuning The WAVE ANALYSER model test and calibration spectrum analysis DSBSC spectrum...41 spectra of unknown signals...41 practical hints...41 future spectral measurements TUTORIAL QUESTIONS Vol A2, ch 5, rev

2 SPECTRUM ANALYSIS - THE WAVE ANALYSER ACHIEVEMENTS: to examine the basic spectrum analyser model; modelling a WAVE ANALYSER; to consider the effects of non-linearities upon performance. PREREQUISITES: completion of the experiment entitled Product demodulation - synchronous and asynchronous in Volume A1. EXTRA MODULES: the SPECTRUM UTILITIES module (not in the TIMS Basic Set of modules). PREPARATION the SPECTRUM ANALYSER Spectrum analysers are found in most homes. They are the domestic radio receiver and TV set. These devices are capable of examining parts of the radio spectrum, and they report what they find either aurally or visually. From their front panel controls we can deduce the frequency from the channel to which they are tuned, and the nature of the spectrum within that channel from the sound and/or picture. A professional spectrum analyser is an instrument for identifying the amplitude spectrum of an electrical signal (typically one whose spectrum does not vary with time). The majority of commercially available instruments cover a very wide frequency spectrum (Hz to GHz), are extremely accurate, and expensive. They generally provide a visual display of the amplitude-frequency spectrum. principle of operation The principle of a spectrum analyser is represented by a tuneable filter, as shown in Figure A2

3 Figure 1: principle of the spectrum analyser. The arrow through the bandpass filter (BPF) shown in Figure 1 implies that the centre frequency to which it is tuned may be changed. The filter bandwidth will determine the frequency resolution of the instrument. The internal noise generated in the circuitry, and the gain of the amplifier, will set a limit to its sensitivity. The symbol of circle-plus-central-arrow represents a voltage indicator of some sort. For the moment we will disregard its response characteristic (RMS, peak, average?), but agree that it will indicate in some way the presence of an output from the filter. The frequency of the signal to which the analyser responds is that of the centre frequency of the BPF. Instruments which require the user to make a manual search, one component at a time, are generally called wave analysers; those which perform the frequency sweep automatically and show the complete amplitude-frequency response on some sort of visual display are called spectrum analysers. practical variation Tuneable bandpass filters are difficult to manufacture. Thus the arrangement of Figure 1 is not used in an instrument covering a wide frequency range. Figure 2 shows a practical compromise. Although this circuit behaves as a tuneable bandpass filter, it uses a fixed lowpass filter. Figure 2: practical spectrum (wave) analyser The arrangement of Figure 1 is a tuned radio frequency receiver (TRF), and that of Figure 2 is based on the principle of the superheterodyne receiver ( superhet ). Refer to the literature circa 1920 to learn about the historical development of these two configurations. The practical difficulties of the former, and the advantages of the latter, are discussed. A2-35

4 The frequency to which the analyser responds is that of the sinusoidal, tuneable, local oscillator. the WAVE ANALYSER This experiment will be concerned with a wave analyser, which was defined above. the model We would like to model a wave analyser that would be of use for future experiments with TIMS. Its tuning range must cover the audio spectrum from say 300 Hz to 10 khz, as well as say 10 khz either side of our standard carrier frequency of 100 khz. Its frequency resolution requirements are modest, determined principally by the fact that we would like to examine the individual spectral components in the sidebands either side of 100 khz modulated signals. In TIMS these are seldom closer than say 250 Hz. A resolution of 100 Hz would be adequate; this is a very modest requirement. We cannot model the arrangement of Figure 1, since we do not have a tuneable BPF with a bandwidth of 100 Hz, covering such a wide range as 250 Hz to 110 khz. The problems associated with the realization of the scheme of Figure 1 are now apparent. The scheme of Figure 2, meeting the above specification, would require a LPF with a cut-off of around 50 Hz. In addition a tuneable oscillator is required; this will need to cover the audio as well as the 100 khz range. TIMS provides the tuneable oscillator in the form of the VCO module. Although a 50 Hz lowpass filter is not difficult to design, there is no such electronic filter in the TIMS BASIC SET of modules. But a moving coil volt meter will serve as the output indicator. Due to the inertia of the mechanical movement it will only respond to DC and very low-frequency signals. It will therefore also serve as the lowpass filter. The SPECTRUM UTILITIES module has been designed for the purpose. spectrum measurement - single component input To make a spectral component measurement it is necessary to understand the principle of the analyser. For a simple first example, suppose the signal v(t) appears at the input, where: v(t) = V 1 cos(2πf 1 t)... 1 and that: VCO output = V vco.cos(2πf 2 t) A2

5 Then: multiplier output = ½.k.V 1.V vco [cos2π(f 1 - f 2 )t + cos2π(f 1 + f 2 )t]... 3 where k is a constant of the multiplier. The LPF filter built into the meter-module will remove the term at frequency (f 1 + f 2 ), and the meter will respond to the term at frequency (f 1 - f 2 ) = δf. Let the amplitude of this signal be V m. V m =... 4 (½.k.V vco ) V 1 Since the amplitude of the VCO output is a constant, the magnitude of the meter reading V m will be proportional to the amplitude of the input component V 1. We will call the constant of proportionality S, the conversion sensitivity 1, so we have: S = 2/(k.V vco.)... 5 V 1 = S.V m... 6 Since V 1 is the amplitude of the unknown signal, this last equation gives the scaling factor to be applied to the meter reading. The frequency of the input component must lie within ± δf Hz of the VCO frequency f 2. The inertia of the moving coil meter prevents it responding to signals of more than a few Hz. For this the VCO frequency must be set close to the frequency of the unknown component at the input. As the frequency difference δf is slowly reduced to zero, the meter will at first quiver (say δf is 10 Hz or less); then start to oscillate with greater and greater swings as δf approaches zero. The peak amplitude of the swing will be V m, reached as δf approaches zero. Despite the last statement, setting the frequency error to precisely zero is not desirable. Should δf = 0 then the term of interest becomes a constant DC voltage, and its amplitude would depend upon the phase angle between the unknown component at the input, and the VCO signal. This phase is unknown, and so would introduce an unnecessary complication. So to measure the amplitude of the unknown component we set δf to one or two Hz, and make a note of the peak reading of the meter and the frequency of the VCO. From this, and the last equation, the unknown amplitude V 1 can be derived. spectral measurement - two component input Suppose there are two components at the input. Provided they are separated by at least the frequency resolution of the analyser, only one will produce an output from the filter at any time 2. 1 conversion, because of the frequency change between input and output 2 this assumes linear operation of all circuits A2-37

6 practical considerations precautions A moving coil volt meter will not respond to signals of more than 10 Hz or so, due to its mechanical inertia. This does not prevent its moving coil from being burned out by other AC signals of excessive amplitude. So, as a practical precaution, the meter in the SPECTRUM UTILITIES module is protected by a low-order LPF. This will also remove the component(s) from the MULTIPLIER at the sum frequency (f 1 + f 2 ). There is a sample-and-hold facility, for capturing the peak swing of the meter. This should be used with care, and its reading not mis-interpreted, since it bypasses the filtering effect of the mechanical inertia of the meter, and will capture all and any signals which reach the meter. Consequently you should have some idea of the relative amplitudes and location of components before using this facility, and an appreciation of the response of the built-in LPF. For further information refer to the TIMS User Manual. searching methods Searching for spectral components takes a certain amount of practice. If the VCO frequency is changed at too great a rate the meter will not have time to respond, and components of significant amplitude will be missed. If and when the meter does respond, adjust the VCO frequency carefully until the meter is oscillating very slowly, and record the peak meter reading. Use the sample-and-hold facility if appropriate. where? In practice one usually has a good idea of where the unknowns are going to be - what is sought is their relative amplitude. Thus the searching process is not as difficult as it might at first appear. how large? No great significance is placed on the measurement of absolute amplitudes - relative amplitudes are what we really want. So pre-calibration is seldom necessary. It is often convenient to tune to the largest component of interest, and then to adjust the meter to full scale deflection (FSD) using the on-board variable resistor RV1, labelled GAIN. This reading becomes the reference A2

7 EXPERIMENT SPECTRUM UTILITIES module The SPECTRUM UTILITIES module contains a centre-reading moving coil meter, with lowpass filtering, and a sample-and-hold facility. Read about it in the TIMS User Manual. Pay particular attention to the precautions necessary if you use the sample-and-hold facility. VCO fine tuning Read about the VCO module in the TIMS User Manual. In the present application it is important to know the techniques of coarse and fine tuning. coarse tuning is done with the front panel f o control (typically with no input connected to V in ). for fine tuning it is convenient to set the GAIN control of the VCO to some small value. Then fine tuning is then done by varying the DC voltage, from the VARIABLE DC module, which is conected to the V in input. The smaller the GAIN setting the finer is the tuning. The WAVE ANALYSER model A wave analyser is shown modelled in Figure 3 below. IN VCO fine tune Figure 3: the WAVE ANALYSER model T1 patch up a model of the block diagram of Figure 2. A suggested scheme is illustrated in Figure 3. Before plugging in the VCO set the on-board switch SW2 to select the VCO mode. Remember the LPF is simulated by the inertia of the moving coil meter movement in the SPECTRUM UTILITIES module. A2-39

8 test and calibration Before using the WAVE ANALYSER to measure some unknown signals it needs to be tested and calibrated in both the audio and 100 khz regions. You can use the output from an AUDIO OSCILLATOR as a source of test signal for the former (say at 1 khz and 10 khz)), and the 100 khz sinewave from the MASTER SIGNALS module for the latter. T2 select a 1 khz sinewave from an AUDIO OSCILLATOR as your source of test signal. Measure its amplitude V 1 at the input to the WAVE ANALYSER, using the oscilloscope. T3 connect the VCO output to the FREQUENCY COUNTER, and tune the VCO to the expected vicinity of the test signal, until the volt meter reading oscillates slowly. Record the peak reading V m of the meter. There is a variable SCALING resistor RV1 on the circuit board of the SPECTRUM UTILITIES module. You may find it convenient, when measuring spectra, to adjust this so the meter reads full scale deflection (FSD) on a reference component - typically the largest to be encountered. T4 check the operation of the on-board SCALING adjustment of the SPECTRUM UTILITIES module. T5 calculate the conversion sensitivity V 1 / V m of your WAVE ANALYSER. T6 repeat the last three tasks for a 100 khz test input. You will now have three determinations of the sensitivity S of the WAVE ANALYSER. Ideally they should all be the same. This assumes: the VCO output amplitude is constant over the full LO and HI frequency ranges the k factor of the MULTIPLIER is independent of frequency probably the k factor of the MULTIPLIER will vary slightly between the LO and HI frequency ranges, and so you may need both an S LO and an S HI. T7 derive an expression for the 1 khz conversion sensitivity of the WAVE ANALYSER in terms of its circuit constants (including the VCO output voltage, multiplier k factor, meter sensitivity). If you do not know the value of k, then set up a temporary arrangement, and measure it. Compare this conversion sensitivity with the direct measurement A2

9 spectrum analysis It is now time to use the WAVE ANALYSER to examine the spectrum of a DSBSC signal - this was promised in the experiment entitled DSBSC generation in Volume A1. DSBSC spectrum T8 set up a DSBSC signal using an AUDIO OSCILLATOR, MULTIPLIER, and the 100 khz sine wave from the MASTER SIGNALS module. T9 use the oscilloscope to measure the amplitude of the DSBSC (in the time domain). From this, and a knowledge of the frequency of the AUDIO OSCILLATOR, sketch the amplitude spectrum of the DSBSC (in the frequency domain). Show clearly the amplitude and frequency scales. T10 connect the DSBSC to your WAVE ANALYSER, and search for spectral components in the range 90 khz to 110 khz. Sketch the measured amplitude spectrum. Show clearly the amplitude and frequency scales. T11 compare the last two spectra, and account for any discrepancies. spectra of unknown signals When happy with the results of the DSBSC spectrum measurement, have a look at the signals at TRUNKS. These have been sent to you for analysis. practical hints In practice it is relative amplitudes which are of interest. Thus one seldom needs to carry out an amplitude calibration. This saves time in setting up the model. Also, one usually knows where the components are, so searching is simplified. It is convenient to find the largest component, and then to set the sensitivity of the meter (with the on-board SCALING adjustment) to indicate full scale on this component. T12 use your WAVE ANALYSER to determine the amplitude/frequency spectra of the signals at TRUNKS. Note that the sensitivity of the SPECTRUM UTILITIES meter can be adjusted with an on-board control. future spectral measurements In future experiments you will find this inexpensive WAVE ANALYSER is a useful measurement tool. The model just examined will be found adequate for your purposes. But remember: its circuits must never be overloaded. In practice this A2-41

10 means that the signal at the analyser input must not overload the input MULTIPLIER. Overload - meaning peak amplitude overload - will mean spurious readings. Use the oscilloscope to ensure that the peak amplitude of the input signal never exceeds the TIMS ANALOG REFERENCE LEVEL. See the Tutorial Question below. TUTORIAL QUESTIONS Q1 the resolution of a wave analyser relates to the width of the window through which it looks at the input spectrum. If the BPF of Figure 1 is brick wall with a passband 2 Hz wide, how would you describe the frequency resolution of the instrument? (note: a filter response even approaching this would be difficult to realize in analog form). Q2 if the LPF of Figure 2 is brick wall, and passes frequencies from DC to 1 Hz, how would you describe the frequency resolution of the instrument? (note: a reasonable approximation to this filter response, in analog form, would not be impossible to implement). In answering the above two questions it was assumed the circuits were linear. Any non-linearity in the circuitry can degrade performance, as is illustrated by answering the next question. Q3 what happens if the amplitude of the input signal is too high? Suppose that there is an amplifier between the input terminal and the BPF of Figure 1, which is brick wall, with a 2 Hz bandwidth. Suppose the amplifier has a non-linear gain characteristic given by: e = a e + a e out 1 in 3 in 3 where a 1 = 1 and a 3 = -0.1 and the input signal e in is a DSBSC. Derive an expression for the spectrum reported by the output meter, in the vicinity of 100 khz, when the input is a DSBSC on a 100 khz carrier, and derived from a 1 khz sinusoidal message. The above question is an exercise in trigonometry. It will illustrate one of the phenomena of intermodulation distortion. hint: give the DSBSC an amplitude V, and look for sum and/or difference ( intermodulation ) components in the vicinity of the sidebands. Show that these are related in a non-linear way to V, and discuss the consequences. Q4 explain the reason for the precautions necessary when using the sample-andhold facility of the SPECTRUM UTILITY module A2